Red Hat, VMware and Eucalyptus in cloud boast bluff-off
Openness is a matter of need
Open... and Shut No matter how you define it, cloud computing is big. The 451 Group forecasts the cloud computing market to hit $16.7bn in revenue by 2013, while Forrester more aggressively projects it to top $241bn by 2020.
A variety of vendors are angling to own a significant chunk of that revenue, with "open" displacing "cheap" as the preferred rallying cry. Among the open cloud proponents, Red Hat, VMware, and Eucalyptus Systems stand out for their very different approaches to, and definitions of, the open cloud.
Prior to cloud computing, we largely defined "open" as a function of source code licensing, with holy wars regularly erupting over how closely vendors or projects actually hew to the spirit and letter of the Open Source Initiative. Measuring the openness of source code alone was never a great measure of true openness, but in the cloud it becomes even less useful.
To determine just how open a cloud computing vendor is, it's still important to look at underlying source code, but APIs and data also factor in – sometimes more prominently than source code.
Red Hat is also quick to point to strong, multi-vendor communities and a lack of patent encumbrances as critical components of truly open clouds, while VMware stresses the importance of open standards, and Eucalyptus insists that adherence to the industry's dominant standard - the Amazon API - factors into any conversation about openness.
So who correctly claims the openness crown? It depends.
While it might be easy enough to develop a cloud from scratch that ticks all the boxes above, it becomes much harder to achieve given the unfortunate fact that enterprises are already burdened by years or even decades of existing, heterogenous infrastructure.
Much of that infrastructure comes from VMware, which dominates the server virtualisation market with its vSphere product. It's no surprise, then, that a big component of VMware's open-cloud strategy is to extend this dominance by offering cloud management tools (vCloud Director) on top of existing vSphere deployments.
Red Hat's general manager of cloud computing, Scott Crenshaw, ridicules the notion that VMware claims to be open, arguing: "VMware will be open the day they open-source vSphere." Which, of course, is approximately "never." And some would likely add that open-source efforts like VMware's CloudFoundry are nothing more than attempts to offer free Platform-as-a-Service (PaaS) bridges to VMware's proprietary and expensive product offerings.
Of course, it goes both ways. Red Hat can claim openness all day long, but it's also the case that Red Hat is pretty restrictive in terms of which hypervisors (no Xen, thank you very much) and operating systems (Want CentOS? Better look elsewhere) it supports. Not that VMware supports all hypervisors, either.
But then, there's a lot of pots calling kettles black in cloud computing, with these and other vendors. And it's mostly harmless.
But is it helpful? After all, VMware's strategy of maximising compatibility with existing infrastructure – and then adding things like CloudFoundry so that enterprises can start from scratch on greenfield opportunities – is a hit with CIOs encumbered by existing infrastructure investments. They may not actually care if it's the ultimate in openness. In fact, just as with open source, enterprise CIOs almost certainly don't care.
Red Hat's premise is the world wants heterogeneity. But that's not VMware's experience, according to VMware's senior director of cloud services Matthew Lodge. People want to aggregate their purchases: they want the ability to run their application well. Too much choice and costs increase. Choice becomes a burden.
True, enterprises also don't want to be locked into a product-level decision for the rest of their lives, but VMware offers options to get out. Many cloud providers – basically, anyone an enterprise would want to use – have built clouds on top of VMware's cloud stack, and each must support open virtualisation file formats and the vCloud API, which requires import/export of data in and out of the cloud). Oh, and that vCloud API? VMware released under the GPL.
But there's more to this story.
In an interview, Red Hat's senior director of product management for cloud Bryan Che pointed out that while VMware can offer incremental benefits to existing vSphere users, most enterprises have a lot of existing software and cloud services that don't come from VMware. Red Hat's strategy, therefore, is an open, hybrid approach that provides an orchestration layer that lets enterprises take their pick of public cloud providers, virtualisation technologies, etc. It's something CIOs can incrementally layer on top of whatever they already have within their enterprise.
The downside to Red Hat's approach is its sheer complexity. With such a diversity of systems to support, no single vendor can hope to support it all. But this is where open source and community come in, two areas in which Red Hat has extensive experience.
DeltaCloud is a prominent example. Red Hat wrote and donated the DeltaCloud code to the Apache Software Foundation (now a top-level project within ASF), and has helped to foster it as a community-driven project in which Red Hat is an active participant, but not the owner. Kind of like Linux.
Red Hat then goes one step further to ensure that enterprise data is easily portable. As Che told me, if enterprise data is stuck at Amazon, even if I can bring the app over to VMware and run it in the data centre, a different kind of openness is necessary to avoid lock-in. Red Hat acquired Gluster to enable it to provide enterprises a global namespace so that they can have uniform access to their data wherever it is.
That's pretty open.
But it's not the only definition of open that matters. As noted, VMware's virtualisation software currently dominates the data centre, and Amazon dominates public cloud computing. So any cloud story that doesn't embrace these giants is a losing strategy.
Which is why Eucalyptus is so interesting. Eucalyptus, like Red Hat, actually protects an enterprise's existing VMware investment by orchestrating existing VMware virtualisation environments such as ESX and ESXi, then offers freedom for the future. Eucalyptus supports all major hypervisors and supports the de facto Amazon API standard, neither of which is true of VMware, and in the case of Red Hat, it can't match Eucalyptus' support for the Amazon API.
In other words, Red Hat, VMware, and Eucalyptus are all open, but in different ways. A CIO is going to find much to love in each approach, though each also comes with built-in shortcomings. Red Hat and Eucalyptus have more traditional "openness" bragging rights, and Red Hat has a long track record of talking the openness talk and walking the openness walk. But VMware is taking significant steps to open itself up in ways that make its data centre dominance less concerning to CIOs inclined to also try it out in the cloud.
All of which means that CIOs probably aren't going to be able to pick a cloud vendor purely by answering the question: "Which is the most open?" That's a question with all sorts of different, correct responses.
The better question is: "Which will help a CIO make sense of existing infrastructure and provide for the future?" Each of these three vendors has a good answer to that question, making the answer a function of an enterprise's individual circumstances. ®
Matt Asay is senior vice president of business development at Nodeable, offering systems management for managing and analysing cloud-based data. He was formerly SVP of biz dev at HTML5 start-up Strobe and chief operating officer of Ubuntu commercial operation Canonical. With more than a decade spent in open source, Asay served as Alfresco's general manager for the Americas and vice president of business development, and he helped put Novell on its open source track. Asay is an emeritus board member of the Open Source Initiative (OSI). His column, Open...and Shut, appears three times a week on The Register.
Sponsored: Virtualization security practical guide